Tetrahedral geometry attention for GAIA benchmarks - 64-point synergetic transformer
-
Updated
Jan 14, 2026 - Python
Tetrahedral geometry attention for GAIA benchmarks - 64-point synergetic transformer
A general-purpose AI agent powered by Claude and the Agent SDK — web search, file processing, vision, audio transcription, precise math, and multi-agent orchestration from your terminal.
Final assignment for Hugging Face's Agents Course (Unit 4) — a GAIA benchmark agent built with smolagents CodeAgent + GPT-4o, deployed on Hugging Face Spaces to answer and auto-submit Level-1 GAIA validation questions via their scoring API.
Streamlit tool for benchmarking OpenAI models against GAIA test cases with feedback-driven re-evaluation
Repository with a LangGraph general purpose AI Agent for final assignment of HuggingFace Agents course.
PECS: 基于 LangGraph 的四角色多智能体任务求解框架(Planner/Executor/Critic/Synthesizer)。WebShop 真实环境 +25pp (25% vs 0%);GAIA 官方 53 题 26.4% vs ReAct 24.5%(McNemar 不显著)。含 AST 沙箱、50000 token 硬预算、FastAPI 限流/混沌/CI/Prometheus。
🕵🏻♂️ LangGraph-powered AI agent tackling GAIA Benchmark tasks, built for Hugging Face Agents Course.
Supervisor multi-agent system for autonomous research, reasoning and code execution. Built on LangGraph. Targeting GAIA.
To associate your repository with the gaia-benchmark topic, visit your repo's landing page and select "manage topics."